JP2008139924A - Memory stock system - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a memory stock system which can be reasonably used even by an elderly person. <P>SOLUTION: This memory stock system is provided with a means for making a question to derive the memory of a user; a means for receiving an answer to the question and a means for storing the answer. A means for making a question and answering the question by turns in an interactive format, and for outputting the answer is configured as a loudspeaker or a display device, and the means for receiving the answer is configured as a microphone or a keyboard, and the means for storing the answer is configured to be separable from the memory stock system. <P>COPYRIGHT: (C)2008,JPO&INPIT

Description

  The present invention relates to a system for taking out all the contents stored by human beings as faithfully as possible and reversibly stocking them in an electronic device.

In Patent Document 1, a site is created on the Internet homepage to store various information such as the member's own history creation and memories. Based on the request of the member, all or a part of the recorded information is browsed and specified. Contents that can be provided to the public, released to the public, and designated for those times are disclosed.
Japanese Patent Application Laid-Open No. 2004-228561 discloses a game machine including a means for asking a user a question and a voice recognition means for receiving an answer to the question.
JP 2002-197287 A Special table 2003-526833

Human memory fades over time and eventually disappears completely. In addition, the range of thinking becomes narrower with age, causing Alzheimer-type senile dementia.
Also, accidents and illness can cause memory loss.

  In such a case, if the memory is backed up, it can be reproduced by the development of medicine and science. In addition, even after the brain has stopped, as long as the person's memory remains, a medium that replaces the brain in which the memory is implanted will be developed, so that the person who has died can be revived. .

  As described in Patent Document 1, even if you can record things such as your own history and memories that you know well, there is some memory that is fading or that is potentially stored but not revealed You can't record without a chance.

  In addition, in order to perform the above-described reproduction and restoration, it is necessary to store a huge amount of memory accurately for each individual, for example, by adding the type and strength of the memory. Patent document 1 discloses that one's own history is recorded, but one's own history was born in XX year, entered XX university in XX year, married in XX year, etc. It is just a record of the facts that have occurred and is just a part of the record without any emotion. Memories are fragmented, and will not enable memorization and reproduction at a later date.

  Moreover, even if the contents disclosed in Patent Document 2 are combined with Patent Document 1, it is convenient to create one's own history, but it does not record all of the current memory contents in one's own brain. Because it is an irreversible storage means, it cannot be a technology that can be played back or revived.

  In order to solve the above-mentioned problems, a storage stock system according to the present invention has means for issuing a question for forcibly extracting a user's memory, means for receiving an answer to the question, and reversibly storing the answer. Means are provided, and the generation of the question and the storage of the answer are alternately performed according to the question-answer format.

  Since the answer is reversibly tagged and stored, if a search is performed later on the means for storing the tag as a keyword, a list of related answers can be generated from the means for storing contrary to the time of recording, and If time data is included in the tag, time-sequential layout organization is also possible. Therefore, the same reproduction as the recording at the time of recording by the person who stored the memory in the stock system is possible.

  The means for issuing the question may be a speaker or a display device, and the means for receiving the answer may be a microphone or a keyboard, but is not limited thereto.

  The means for storing the answer may be separable from the storage stock system, and the user's three-dimensional image data and genetic data may be recorded. Making it separable makes it easy to store, and if 3D image data and genetic data are recorded, it can be used as a guideline to create a storage medium (brain and body) for later reproduction and restoration. Become.

  In addition, after the answers to the questions have been completed, information close to the actual memory can be recorded by attaching the kind and weight of emotion to the obtained individual memories. For example, how many emotions of joy, anger, sadness, usually laughter and excitement are included in one memory, how often the phrase used in the answer is used, and how much is required for the answer. This is done according to the length of time and weighting based on the user's own judgment. In addition, the feelings for individual memories may include elements other than those described above, such as likability, disgust, surprise and fear.

  In the above, the emotion and contrast of the voice detected by the microphone can be used to give the emotion type and the degree of lightness, but it is also possible to use a sensor that detects the degree of body temperature, sweating, etc. as an option. It is also possible to parameterize facial expressions using a camera and attach them to tags.

  In addition, the question may be created exclusively for the storage stock system, but based on the words in the dictionary or dictionary (encyclopedia), the words in the dictionary or dictionary from the user's answers A word that has not been asked yet may be selected and used as the next question.

  The question may be given a priority on the words related to the family, and a word that is related to the family and has not been asked yet is selected from the answers of the user as the next question. In this way, questions and answers can be spoken continuously smoothly.

  In the above, as long as the question and the answer are semantically continuously performed smoothly for the user, the content and order of the question may follow any rule.

  According to the present invention, since it is performed in a question-and-answer format, the user does not know from what to answer, and all or a memory close to it can be retrieved smoothly. Also, unlike forms like filling in your own history, you will think and answer unforeseen questions yourself, forcibly revealing the potential memory pushed into one corner of the brain. It will be.

  Furthermore, since the memory is stored by stock means that can be reversibly input / output, if the extraction of the memory is carried out in a nearly perfect state, the memory lost by the accident can be restored, the reproduction after brain death, that is, the recovery after death. I can expect.

  Embodiments of the present invention will be described below with reference to the accompanying drawings. FIG. 1 is a functional block diagram of a storage stock system according to the present invention using a game machine, FIG. 2 is an explanatory diagram of operation of a neural network circuit and a database, and FIGS. 3 (a) and 3 (b) show the structure of the neural network system. The schematic diagram shown is shown.

  FIG. 1 shows a functional block diagram of a storage stock system according to the present invention using a game machine. The storage stock system 1 includes a microphone 2 and a speaker 3 as input / output functions connected to the game machine 6, a neural network circuit 7 for neural network processing, a general-purpose database 8, and a personal database. The personal database 9 stores the answer of the user 4 to the question generated by the storage stock system 1 and the personal data generated from the answer.

  The microphone 2 and the speaker 3 may be built in the game machine 6 or may be connected to the game machine 1 by wire or wireless.

  When the user 4 outputs a question generated by the storage stock system 1 using data in the general-purpose database 8 or the personal database 9 from the speaker 3, the user 4 answers the heard question to the microphone 2 by voice. .

  The voice is converted into a character data group by a voice recognition function inside the storage stock system 1, and the character data group, together with the original character data, is a keyword generated by the neural network from the character data group as a search index. It is stored in the personal database 9 of the storage stock system 1.

  The general-purpose database 8 stores general knowledge data such as, for example, Kojien (registered trademark), the Daiwa Japanese dictionary of the research company, the Daiwa English dictionary, and the basic knowledge (registered trademark) of modern terms.

  Since general knowledge data stored in the general-purpose database 8 needs to be prepared according to the native language of the user 4, a plurality of general-purpose databases 8 are provided for the user 4 who operates a plurality of languages. In this case, the general-purpose database 8 may be placed in an external network such as the Internet.

In the above, a dedicated device may be used without using a game machine. Further, not the text but the voice itself may be stored in the database together with the keyword. When storing voice, the stored voice can be output as it is, and in the case of sentences, in addition to the screen display, synthesized voice can be pronounced. This is useful when playing the 4th voice.

Furthermore, the keyword and the character data group or the voice data may be stored in different databases with a correlated index. In this case, for example, a keyword with a small amount of data is stored in the memory of the game machine main body or the game cartridge, and a character data group or a sound data with a large amount of data is a file that is common to different mass storage devices that are individually recognized. It is conceivable that a number is assigned and stored.

  In the above, the data processed by the neural network circuit 7 is weighted according to the appearance frequency, formatted, then categorized, and stored in the personal database 9 with attribute data such as emotion added. However, the data stored in the personal database 9 is used as data to be input to the brain that has lost memory due to an accident or the like, or as data to be input to the medium when a medium that replaces the brain is developed. be able to.

In order to faithfully reproduce the user's memory, it is necessary to record the connection of each memory, the lightness of the memory, the presence / absence of emotions and emotions of each memory, and their strength. These elements are different for each individual user, and this difference is the individuality that distinguishes each user from others. In order to realize this, the neural network circuit 7 is optimal, and the reproducibility varies greatly depending on how the attributes are classified.

In the above, the personal database 9 of the game machine has a structure in which a removable storage means such as a memory stick (registered trademark), a USB memory, an SD card, etc. can be inserted and removed, and answers to questions in the game can be obtained. It is conceivable that the means for storing is separable from the storage stock system.

Further, when the personal database 9 is a removable storage means, it is desirable that a personal authentication data is disclosed only to registered persons including the person himself / herself by using a fingerprint authentication device or the like together.

FIG. 2 is a conceptual diagram of a hierarchical neural network. This is a multi-module neural network composed of a basic network that performs local information processing and a connected neural network that exchanges processed information.

In an actual neural network, the data moves complicatedly by regression analysis, etc., but this diagram is modeled for clarity. In FIG. 2, the general-purpose database 8 is connected to the connected neural network 7 b and the individual database 9 is connected to the connected neural network 13. In operation, the general-purpose database 8 and the personal database 9 can be accessed from other neural networks as necessary.

  The functions of the neural network circuit 7 include a basic neural network 11 for modeling input data so that it can be easily processed by a subsequent neural network, and a connected neural network 12 for classifying the modeled data into a predetermined category. ~ 14.

  For the basic neural network 11, for example, a Kohonen self-organizing feature map may be used. For the connected neural networks 12 to 14, learning of a hierarchical neural network by backpropagation can be considered.

  The connected neural network 11 is configured by connecting a plurality of neuron models, and is classified into a multi-module neural network using associative memory by combining hierarchical networks.

  A multi-module neural network performs data mining in a neural network that exchanges processed information, for example, signals from each input from a basic network not only flow unilaterally to a plurality of neurons in a connected network, but also Associative memory is realized by learning various combinations in the connected network while being fed back to other input terminals of the connected network.

  In addition, by using a self-propagating neural network, when the number of input data that best matches one neuron exceeds a certain number, the neuron is split and divided into two groups, and the number of input data per group It is also conceivable to reduce the variation of.

  The input data 10 is output by the output unit 15 as data to which the classification attribute is added by the connected networks 12 to 14 after being divided by the template represented by the category attribute by being filtered in the basic network 11. This is because, after the input data is largely classified by the basic network 11 composed of the vector quantization neural network, the connected networks 12 to 14 which are interconnected networks with feedback having the function of finely classifying the data are used to answer the question. Depending on the content of the data, appropriate feedback is performed to improve the accuracy of the memory.

  By having the structure of the basic neural network 11 and the three-stage connected neural networks 12 to 14 or the three-stage or higher connected neural network as described above, an object-oriented interactive database is generated from the input 10 and used as the database. Formatted data that can be easily searched can be output from the output unit 15. As a result, it is easy to assign and classify the input memory (user's answer) as an easily searchable type.

  In the modeling function using the first basic neural network 11, for example, a sentence of a character data group is analyzed according to the 5W1H rule, and if the desired data is not included in the character data group, The character data group is searched to collect 5W1H data.

  Next, in the connected neural network 12 connected to the general-purpose database 8 by the data line 8a, the general-purpose data is detected from the text of the character data group, and is generated and corrected by adding or correcting the data. Make index data versatile. For example, if the text is a group of character data such as “I lived alone in Tokyo in the year of the Tokyo Olympics,” this year was 1964 using a general-purpose database 8 with chronological data. Therefore, 1964 time data is added to the index. The purpose of the reversible storage stock system according to the present invention is to enable the user 4's memory at the time of input to be replayed at a later date, so the time data of 1964 is only used for searching, If the user 4 does not remember the year 1964, it is not necessary to add the data 1964 when restoring the data.

  In the connected neural network 13, in the connected neural network 12, in addition to the text of the character data group to which the data of the general-purpose database 8 input from the data line 8a is added, the personal database 9 connected by the data line 9a. By using this data, it is possible to perform relative processing in the knowledge of the user 4. For example, when a sentence of a character data group that “a dog kept at home was a white dog” is input, the stored data stored in the personal database 9 is “the dog at home is black. If the character data group “I couldn't see clearly at night” was found, it is difficult to determine whether the dog in the house is black or white, so this memory can be attributed as an “ambiguous memory”. As described above, the purpose of the reversible storage stock system according to the present invention is to allow the user 4's memory at the time of input to be replayed at a later date, so that ambiguous memory is recorded as ambiguous memory. It is a feature of the reversible storage stock system of the invention.

  In the connected neural network 14, the sentence of the character data group normalized using the data of the general-purpose database 8 and the personal database 9 is extracted by using the neural network algorithm and regression analysis to extract the attribute. A text data group text file is output from the output data line 16 to the personal database 9 and registered in a predetermined category.

  The connected neural network 14 is used for memory weighting, and when a specific word or expression is frequently used directly or indirectly by the user 4 by referring to a lot of data, the specific word is used. Set the degree of weighting in expressions and expressions high. In this application, the hierarchical neural network based on the back propagation is an effective means.

  For example, even if you ask the same question, if there is a lot of data related to a specific thing (for example, a mother or father) in the contents of the answer, it is judged that you have a deep feeling for that particular thing, To increase the weight.

  In the personal database 9 connected to the output unit 15 and the output data line 16, the user's answer to the question is registered and stored.

  It is also conceivable to add a personal profile to the personal database 9. As the detailed data of the personal profile, biometric information for personal identification, personal history, family tree, and the like can be considered.

  The index data registered in the personal database 9 has a role of FAT (file allocation table) used in a hard disk or a floppy (registered trademark) disk. In addition to the 5W1H data, the index data may be encoded or indexed with six emotions of joy, anger, sadness, laughter, and excitement as memory emotions. It is conceivable that the memory is weighted, and the light weight is determined by the frequency of use of the phrase used in the answer, the length of time required for the answer, or the weight based on the user's own judgment.

  The biometric information for personal authentication stored in the personal database 9 includes fingerprint information, foot fingerprint information, palm print information, iris pattern information, face shape information, voice feature information, signature information, and deoxyribonucleic acid (DNA) information. It is conceivable to include at least one of This is a certificate of authenticity of whether the personal database 9 is authentic. In order to verify the authenticity of the data, for example, a copyright management system using legal expressions described in JP-T-2006-501536 can be used.

  In the above description, the biometric authentication information need not be single data, and the biometric authentication information may be composed of a plurality of data. In this case, it is desirable that at least DNA data is included as biometric authentication information in order to prove that the data is the person's data even after the person's death.

  As DNA data, the human genome DNA information itself consisting of about 3 billion base sequences having a double helix structure is most definitely preserved, and the human genome DNA information is convoluted in the cell, and its sequence It is said that all cells are the same and are said to be unique information that will last forever. However, if it is difficult to save all 3 billion data, it is difficult to secure memory space. Only DNA information of unrelated regions (hereinafter referred to as DNA-ID) may be stored. Further, when the memory mechanism of the brain cell becomes clear in the future, the memory content may be directly stored as data.

  The database in the present invention is preferably object-oriented. The database is an object database that stores data obtained interactively and generates a database. The database is preferably a network type rather than a relational type or a hierarchical type, and the file structure is also a network type file structure. It can be considered. By using the network type file structure, a certain file can be referred to from a plurality of places, and another file can loop and refer to itself.

  As the database, for example, if it is a relational database, Zope (registered trademark), PostSQL (registered trademark), Oracle (registered trademark), or the like can be used.

  FIG. 3A is a diagram showing a basic structure of a database according to the present invention. One set of data 17 includes an ID header 18 and text metadata 19. The ID header 18 generates a single data set 17 by modeling the body metadata 19 and inserting the keywords A to H into the format. The metafile 19 is data for storing text of a character data group used as input data of the neural network circuit 7 and voice data of the user 4 as a metafile.

  When voice data of the user 4 is used as the metafile 19, all the data obtained by dictation is stored as a metafile. However, since the metafile 19 is directly searched, it takes time. Searching is facilitated by extracting necessary keywords from the file 19 as the ID header 18.

  FIG. 3B is a diagram showing details of the ID header 18 in the basic structure of the database. The ID header 18 has a 5W1H element exemplified by A to H, and has a structure that allows additional writing. Since it is possible to add data, it is not always necessary to have all the data completely aligned. By storing only the elements that can be recognized by listening and leaving blanks for memories that cannot be remembered, the personality can be reproduced at a later date.

  In the above, for the purpose of reproducing the person himself, it is a precondition that the person's memory at the time of the question is left as it is, and it is not a question that can not be answered because it does not recall the memory that is forcibly disappearing The answer is not necessary.

  Further, when the stored database 9 is restored, a method of reproducing the user 4 by using a template and inserting keywords is conceivable. For example, in the self-introduction mode, “My name is (Taro Yamada). I am from (Nagano), and I was born in (July 3, Taisho 14) in (Fushishi Matsumoto City, Nagano Prefecture). My father is (Yamada Ichiro) and my mother is (Yamada Hanako). "Create a sentence by inserting the keywords picked up from the database 9 into the brackets of a general greeting, or write the sentence to a DSP. It is conceivable that voice processing is performed with a high-speed processing device such as (Digital Signal Processor) and output as pseudo personal voice.

  Furthermore, in the random playback mode, when an arbitrary keyword, for example, a keyword “Dan Candy Shop” is input to the revival machine such as the game machine 1 by voice or text, it is registered with the keyword “Dan Candy Shop” in the database 9. It is conceivable that the sentence data or the voice data is extracted and reproduced in accordance with the age or the order of the feelings.

  The memory stock system according to the present invention makes it possible to reproduce the user's own memory as faithfully as possible, and it is possible to provide a personal database with individuality. In the future, DNA regeneration technology will be provided at a sufficient level. In addition, it is possible to virtually reproduce the principal by using the principal data stored in the storage stock system of the present invention.

Functional block diagram of storage stock system Conceptual diagram of hierarchical neural network (A) Diagram showing basic structure of database (b) Diagram showing details of ID header

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 ... Memory stock system 2 ... Microphone 3 ... Speaker 4 ... User 5 ... Voice recognition circuit 6 ... Game machine 7 ... Neural network circuit 8 ... General-purpose data base 8a ... Data line 9 ... Personal data base 9a ... Data line 10 ... Neural Network input data 11 ... Basic neural network 12 ... Connected neural network 13 ... Connected neural network 14 ... Connected neural network 15 ... Output unit 16 ... Output data line 17 ... Data set 18 ... ID header 19 ... Meta file

Claims (7)

A means for issuing a question for extracting a user's memory; a means for receiving an answer to the question; and a means for storing the answer in a reversible manner. The question and the answer are alternately performed according to a question / answer format. A storage stock system characterized by that. 2. The storage stock system according to claim 1, wherein after the answer to the question is completed, the type and weight of emotion are assigned to each memory. 3. The memory stock system according to claim 2, wherein the emotions of the memory are six emotions of joy, anger, sadness, usually laughter and excitement, and the lightness of the memory is the use of the phrase used in the answer. A storage stock system characterized by frequency, length of time required for an answer, or weighting by the user's own judgment. 2. The storage stock system according to claim 1, wherein the question is based on words in a dictionary or dictionary, and words that are in the dictionary or dictionary and are not yet questioned from among the answers of the user. A storage stock system, characterized by selecting and using the word in the next question. 2. The storage stock system according to claim 1, wherein the question is based on a vocabulary obtained by combining the accumulated database and an existing terminology dictionary / encyclopedia, and the database or the terminology dictionary / encyclopedia is selected from answers of users. A memory stock system characterized by selecting words that are listed in the above and that have not been asked yet and using the words in the next question. 2. The storage stock system according to claim 1, wherein said question gives priority to a word related to a family, selects a word related to a family and has not yet been asked from a user's answer, and A memory stock system characterized by using the words in questions. 2. The storage stock system according to claim 1, wherein the means for storing the answer is separable from the storage stock system, and the user's three-dimensional image data and genetic data are recorded incidentally. And storage stock system.

Images